Cohort analysis is a way to track the "density" of a metric for specific user groups. Unlike regular graphs, cohorts show exactly the stickiness of users, distributed over time.
It sounds difficult, but in reality it is easier to understand. Let's take a concrete example: let's try to understand how the subscriber base behaves in a mobile application: how they pay, unsubscribe and how long they live. I will show the answers to these questions on the example of our implementation in Adapty .
When constructing cohorts, it is important to indicate how they are formed, what to display in cells, and what to measure. In our case, cohorts are formed by the month of installation, in the cell we show the revenue, the number of subscribers and the average revenue per subscriber for each month (ARPPU).
, () , , — ( 1). 1 — , 2 — . : , , . :
, . , . CPA (CPI). , , , , , . , , , 2.
, $2900 73 :
: . 67 , 38, 10 ( , , ).
?
, $4000 , « ?», $4000 .
.
( — )
$2900 Apple, $2465 15% ( Apple SMB). , $10.
, 2 , 20%, 10% 24. , — 24 . $15.48 , ARPPU .
$4000, , $1500. $372 , 4-5 . , , , , , , , , . , , , , . .